Koremenos et al. (2001) argued that states design international institutions deliberately and that the variation among them can be described on a small number of dimensions: membership, scope, centralisation of tasks, rules of control and flexibility. Their conjecture that flexibility increases with uncertainty about the state of the world bears directly on water treaties under a changing climate. From the study of common-pool resources, Ostrom (1990) identified features of long-lived institutions, among them monitoring, graduated responses to violations and accessible arenas for resolving conflict. Young (2002) added the problem of fit: an institution performs well only if its rules match the properties of the biophysical system it governs.

The comparative water literature established that shared rivers produce far more cooperation than violence (Wolf, 1998) and that the risk of dispute rises when physical or political change outpaces the capacity of institutions to absorb it (Wolf et al., 2003). Sadoff and Grey (2002) showed that negotiations broaden when parties bargain over the benefits of water and not only its volume. On variability specifically, Fischhendler (2004) examined how treaties adapt legally to climatic uncertainty, and Drieschova et al. (2008) classified the mechanisms treaties use to deal with fluctuating flow, distinguishing fixed from proportional allocation and noting provisions for drought, consultation and amendment. De Stefano et al. (2012) combined projected variability with treaty content to identify basins whose institutions are poorly equipped for what is coming.

Design does not operate in a vacuum. Zeitoun and Warner (2006) introduced hydro-hegemony to describe how the stronger riparian shapes outcomes through its position on the river, its material power and its control of the terms of debate, so that an agreement may record an imbalance as much as resolve one. In South Asia, India is the upper riparian to Pakistan and Bangladesh and the lower riparian to Nepal and China, which makes the region unusually suited to studying how the same state designs institutions from different positions. The legal history of the agreements is set out by Salman and Uprety (2002).

The unit of analysis is a negotiation episode: a defined period in which the parties bargain over the creation, interpretation or revision of an arrangement for a named river. Bounding matters because the same treaty can generate distinct episodes with different dynamics. Under the Indus Waters Treaty, the reference of the Baglihar project to a neutral expert and the arbitration over the Kishenganga project are separate episodes within one institution. The analyst records the parties, the river reach, the opening and closing events, and any third party involved.

The primary sources are the treaty text and annexures, the published decisions of dispute settlement bodies, joint statements and the communiqués of bilateral commissions. Secondary legal and historical accounts supply context (Salman and Uprety, 2002). Because the minutes of bodies such as the Permanent Indus Commission and the Joint Rivers Commission are largely unpublished, the analyst should state which records were available and treat silence in the public record as missing information, not as evidence that nothing occurred.

The framework does not ask the political analyst to model hydrology. It asks for a short, sourced description of how the flow regime relevant to the treaty is expected to change, taken from published assessments (Wester et al., 2019). Three aspects are recorded: change in the timing of flow, change in total volume, and change in variability between years. The description is tied to the treaty's own terms. For an agreement that shares water in a defined dry season window, timing is the aspect that matters; for one that allocates whole rivers, it may be volume and the frequency of extremes.

Six features are coded from the treaty text, each on a simple ordinal scale with the supporting clause quoted. They adapt the dimensions of Koremenos et al. (2001) to water using the mechanisms catalogued by Drieschova et al. (2008). Table 1 lists the features with the question each answers and an illustration from the region.

The standing joint body is coded as a feature in its own right because its mandate cuts across the others: it is usually the channel for data exchange and the first rung of dispute settlement. Coding should be done independently by two readers where possible, with disagreements recorded, since treaty language on matters such as consultation in times of low flow is often open to more than one reading.

A clause that is never invoked tells a different story from one in constant use. For each coded feature the analyst records whether, within the episode, it was invoked, by whom and with what result; whether the parties worked around it through political channels; and whether either party contested its meaning. Here the hydro-hegemony approach supplies the questions: which party controlled the timing of talks, whose data were treated as authoritative, and whether the weaker riparian complied, contested or sought outside support (Zeitoun and Warner, 2006). India's announcement in April 2025 that it was holding the Indus Waters Treaty in abeyance is an example of practice departing from text, since the treaty contains no provision for unilateral suspension.

The final step applies the method of structured, focused comparison described by George and Bennett (2005): the same questions are put to each episode, and the answers are set side by side. With a handful of cases the aim is not statistical inference. It is to see whether episodes with similar stress and different designs, or similar designs and different power relations, unfold differently, and to trace within each case how a design feature affected the course of bargaining. The assessed fit recorded at the base of Figure 1 is a reasoned judgement supported by that tracing.